ORIGINAL RESEARCH article

Front. Big Data

Sec. Machine Learning and Artificial Intelligence

Volume 8 - 2025 | doi: 10.3389/fdata.2025.1557779

A FASHION PRODUCT RECOMMENDATION BASED ON ADAPTIVE VPKNN-NET ALGORITHM WITHOUT FUZZY SIMILAR IMAGE

Provisionally accepted
Sabitha  RSabitha R1*Sundar  DSundar D2
  • 1V.V. Vanniaperumal College for Women,, Virudhunagar, India
  • 2Government Arts College Melur, Madurai, Tamil Nadu, India

The final, formatted version of the article will be published soon.

Recommender systems play a critical role in e-commerce by helping users navigate vast product catalogs and discover relevant items. However, traditional keyword-based systems often fall short in visually driven domains like fashion, where subjective style preferences are difficult to express textually. This research aims to develop a visually aware fashion product recommendation framework using an Adaptive VPKNN-net algorithm. The proposed method integrates deep visual feature extraction via a pre-trained VGG16 Convolutional Neural Network and dimensionality reduction through Principal Component Analysis (PCA). It employs a modified K-Nearest Neighbors (KNN) algorithm that combines Euclidean and cosine similarity metrics to improve visual similarity assessments. Unlike fuzzy logic-based approaches, the proposed method offers a precise and interpretable recommendation mechanism. Experiments conducted on the Fashion Product Images (Small) dataset from Kaggle demonstrate that the Adaptive VPKNN-net model achieves superior performance, with a classification accuracy of 98.69%, and lower RMSE (0.8213) and MAE (0.6045) compared to baseline models. In conclusion, the proposed framework effectively enhances the accuracy, efficiency, and interpretability of fashion product recommendations, offering a scalable solution for visually driven e-commerce platforms with limited training data.

Keywords: Recommendation model, Collaborative Filtering, K-nearest neighbor, Principal Component Analysis, Convolution Neural Network, E-commerce. 1. Introduction

Received: 11 Jan 2025; Accepted: 07 Jul 2025.

Copyright: © 2025 R and D. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

* Correspondence: Sabitha R, V.V. Vanniaperumal College for Women,, Virudhunagar, India

Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.